A off-topic messge: I modified your comment to use "Preformatted" from the Format drop-down box. I would recommend using Preformatted as oppose to Code, or the BBcode [code] tags.
An alternative is to use pastebin.com when posting your log info, and provide a URL.
Unfortunately the part of the log I was thinking about is not one of the ones you posted. I was thinking of the start of the log after a full restart of Boinc or a reboot of the computer where the GPU detection is logged.
pkunk wrote:
This is with the following settings configured on E@H webpage. Which I presume are being overridden by the APP_CONFIG.XML.
Yes, the app_info will override the settings on the website for the apps specified in the app_info.
pkunk wrote:
BRP = 0.25 <-- Should do four (4) tasks at a time correct?
Yes
1.0 = x1
0.5 = x2
0.33 = x3
0.25 = x4
and so on, allowed values are between 0.0 - 1.0.
Thank you for editing it. The first one I came across was the Code formatting. I presume we are not able to edit posts after making them. For longer log files I will host them offsite and put a link to them here. Thank you for the tip.
I should have time tonight or tomorrow to do a shutdown/reboot of the machine and get the start of the log file you asked for. Thanks again for the help.
Well this morning I got up and the machine was frozen so I restarted it and this is the start of the log. It is showing the (BRP=0.25) working correctly however the tasks have slowed to 0.312%/hour. Whereas with two GPU (BRP = 0.5) processes it was doing about 80%/hour
Here is the log file:
10/20/2016 7:15:04 AM | | Starting BOINC client version 7.6.22 for windows_x86_64
10/20/2016 7:15:04 AM | | log flags: file_xfer, sched_ops, task, coproc_debug
10/20/2016 7:15:04 AM | | Libraries: libcurl/7.45.0 OpenSSL/1.0.2d zlib/1.2.8
10/20/2016 7:15:04 AM | | Data directory: C:\boinc\data
10/20/2016 7:15:04 AM | | Running under account rick
10/20/2016 7:15:04 AM | | [coproc] launching child process at C:\boinc\boinc.exe
10/20/2016 7:15:04 AM | | [coproc] relative to directory C:\boinc\data
10/20/2016 7:15:04 AM | | [coproc] with data directory "C:\boinc\data"
10/20/2016 7:15:04 AM | | OpenCL: AMD/ATI GPU 0: Ellesmere (driver version 2117.13 (VM), device version OpenCL 2.0 AMD-APP (2117.13), 8192MB, 8192MB available, 3853 GFLOPS peak)
10/20/2016 7:15:04 AM | | OpenCL CPU: Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 2117.13 (sse2,avx), device version OpenCL 1.2 AMD-APP (2117.13))
10/20/2016 7:15:04 AM | | [coproc] No NVIDIA library found
10/20/2016 7:15:04 AM | | [coproc] No ATI library found.
10/20/2016 7:15:04 AM | | Host name: DESKTOP-LDRLIQH
10/20/2016 7:15:04 AM | | Processor: 32 GenuineIntel Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz [Family 6 Model 45 Stepping 7]
10/20/2016 7:15:04 AM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes syscall nx lm avx vmx smx tm2 dca pbe
10/20/2016 7:15:04 AM | | OS: Microsoft Windows 10: Professional x64 Edition, (10.00.14393.00)
10/20/2016 7:15:04 AM | | Memory: 31.97 GB physical, 33.97 GB virtual
10/20/2016 7:15:04 AM | | Disk: 206.35 GB total, 129.08 GB free
10/20/2016 7:15:04 AM | | Local time is UTC -7 hours
10/20/2016 7:15:04 AM | Einstein@Home | Found app_config.xml
10/20/2016 7:15:04 AM | | Config: don't compute while lotroclient.exe is running
10/20/2016 7:15:04 AM | | Config: don't use GPUs while lotroclient.exe is running
10/20/2016 7:15:04 AM | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 12323896; resource share 100
10/20/2016 7:15:09 AM | Einstein@Home | General prefs: from Einstein@Home (last modified 18-Oct-2016 19:38:10)
10/20/2016 7:15:09 AM | Einstein@Home | Host location: none
10/20/2016 7:15:09 AM | Einstein@Home | General prefs: using your defaults
10/20/2016 7:15:09 AM | | Preferences:
10/20/2016 7:15:09 AM | | max memory usage when active: 16366.97MB
10/20/2016 7:15:09 AM | | max memory usage when idle: 29460.55MB
10/20/2016 7:15:09 AM | | max disk usage: 10.00GB
10/20/2016 7:15:09 AM | | max CPUs used: 4
10/20/2016 7:15:09 AM | | suspend work if non-BOINC CPU load exceeds 40%
10/20/2016 7:15:09 AM | | (to change preferences, visit a project web site or select Preferences in the Manager)
10/20/2016 7:15:09 AM | Einstein@Home | [coproc] Assigning 0.250000 of ATI free instance 0 to p2030.20141130.G193.80-00.11.C.b2s0g0.00000_1632_1
10/20/2016 7:15:09 AM | Einstein@Home | [coproc] Assigning 0.250000 of ATI instance 0 to p2030.20141130.G193.80-00.11.C.b3s0g0.00000_1616_0
10/20/2016 7:15:09 AM | Einstein@Home | [coproc] Assigning 0.250000 of ATI instance 0 to p2030.20141130.G193.68-00.34.C.b0s0g0.00000_160_1
10/20/2016 7:15:09 AM | Einstein@Home | [coproc] Assigning 0.250000 of ATI instance 0 to p2030.20141130.G193.80-00.11.S.b0s0g0.00000_3312_0
10/20/2016 7:15:09 AM | | suspend work if non-BOINC CPU load exceeds 40%
This is the only line in the log I see that could impact performance but it should pause all tasks when something other than Boinc uses the computer, not slow things down. This is probably not the cause.
pkunk wrote:
I do not know where to go from here.
I'm also running out of ideas...
The only other thing that comes to mind is that OpenCL tasks needs more support from the CPU than e.g. CUDA or they slow down considerably, actually it's not so much about more CPU cycles but faster servicing when needed.
When you try to run more than 2 GPU tasks then how many free CPU cores do you have? Or phrasing it differently, how many CPU tasks are running?
Have you tried not running CPU tasks at all when running more than 2 GPU tasks?
If none of the above works and no one else have any ideas then we might have to accept that running 2 GPU tasks at a time is the max for your system and still having good/decent performance.
Unfortunately, I have tried running it with just GPU tasks and I obtain the same results. Slightly less if you count the CPU tasks that are not being hashed.
Currently I have 13% of my cores comitted which is four (4) of the thirty-two (32) available to the system. Sticking with the stock configuration and changing the BRP = 0.5 allows BOINC to run three (3) CPU tasks and two (2) GPU tasks concurrently. Since for AMD cards the system reserves 0.5 CPU per GPU.
Even when I tried setting it up in the app_config.xml file the system never seemed to dedicate more or less than 0.5 CPU per GPU instance.
Does this line have anything to do with it?
10/20/2016 7:15:04 AM | | [coproc] No ATI library found.
10/20/2016 7:15:04 AM | | [coproc] No ATI library found.
Maybe AgentB can shed some light on this.
The more windows i see the darker it gets.
My guess is that error message refers to CAL which i think is no longer supported by AMD (frozen in 2010 i think) if it is crunching tasks it is nothing to worry about.
Pkunk, A off-topic messge: I
)
Pkunk,
A off-topic messge: I modified your comment to use "Preformatted" from the Format drop-down box. I would recommend using Preformatted as oppose to Code, or the BBcode [code] tags.
An alternative is to use pastebin.com when posting your log info, and provide a URL.
Einstein@Home Project
pkunk
)
Unfortunately the part of the log I was thinking about is not one of the ones you posted. I was thinking of the start of the log after a full restart of Boinc or a reboot of the computer where the GPU detection is logged.
Yes, the app_info will override the settings on the website for the apps specified in the app_info.
Yes
1.0 = x1
0.5 = x2
0.33 = x3
0.25 = x4
and so on, allowed values are between 0.0 - 1.0.
pkunk wrote:Here is what I
)
Thanks we also need the GPU detection stage from the logs with that flag set - the first 40 lines should suffice.
Hi Shawn, Thank you for
)
Hi Shawn,
Thank you for editing it. The first one I came across was the Code formatting. I presume we are not able to edit posts after making them. For longer log files I will host them offsite and put a link to them here. Thank you for the tip.
Hi AgentB and Holmis, I
)
Hi AgentB and Holmis,
I should have time tonight or tomorrow to do a shutdown/reboot of the machine and get the start of the log file you asked for. Thanks again for the help.
Here is the startup of the
)
Here is the startup of the log after rebooting this evening.
End
Oddly enough I thought I have configured it to do four (4) GPU tasks by setting BRP = 0.25. Yet it is only running two.
Well this morning I got up
)
Well this morning I got up and the machine was frozen so I restarted it and this is the start of the log. It is showing the (BRP=0.25) working correctly however the tasks have slowed to 0.312%/hour. Whereas with two GPU (BRP = 0.5) processes it was doing about 80%/hour
Here is the log file:
End.
I do not know where to go from here.
pkunk wrote: 10/20/2016
)
This is the only line in the log I see that could impact performance but it should pause all tasks when something other than Boinc uses the computer, not slow things down. This is probably not the cause.
I'm also running out of ideas...
The only other thing that comes to mind is that OpenCL tasks needs more support from the CPU than e.g. CUDA or they slow down considerably, actually it's not so much about more CPU cycles but faster servicing when needed.
When you try to run more than 2 GPU tasks then how many free CPU cores do you have? Or phrasing it differently, how many CPU tasks are running?
Have you tried not running CPU tasks at all when running more than 2 GPU tasks?
If none of the above works and no one else have any ideas then we might have to accept that running 2 GPU tasks at a time is the max for your system and still having good/decent performance.
Hi Holmis, Unfortunately, I
)
Hi Holmis,
Unfortunately, I have tried running it with just GPU tasks and I obtain the same results. Slightly less if you count the CPU tasks that are not being hashed.
Currently I have 13% of my cores comitted which is four (4) of the thirty-two (32) available to the system. Sticking with the stock configuration and changing the BRP = 0.5 allows BOINC to run three (3) CPU tasks and two (2) GPU tasks concurrently. Since for AMD cards the system reserves 0.5 CPU per GPU.
Even when I tried setting it up in the app_config.xml file the system never seemed to dedicate more or less than 0.5 CPU per GPU instance.
Does this line have anything to do with it?
Maybe AgentB can shed some light on this.
Thanks.
pkunk wrote:Does this line
)
The more windows i see the darker it gets.
My guess is that error message refers to CAL which i think is no longer supported by AMD (frozen in 2010 i think) if it is crunching tasks it is nothing to worry about.